Data Science Projects for Freelancers

We are seeking a highly experienced ML/AI Architect to design, build, and scale enterprise-grade AI/ML and GenAI solutions within our Digital Transformation Practice. The role requires strong technical leadership across ML system architecture, LLM-based solution design, MLOps automation, and cloud-native data engineering. The ideal candidate will drive innovation, establish best practices, and deliver production-ready AI solutions at scale

-----------------------------------
| Key Skills & Competencies |
-----------------------------------

LLMs / GenAI: GPT-4/5, Llama 3/3.5, Qwen, Claude, RAG, prompt engineering, LoRA/QLoRA, embeddings
ML / DL: PyTorch/TensorFlow/Transformers/CNNs/XGBoost/NLP/NLU/NER
MLOps: MLflow, Airflow, Kubeflow, SageMaker, Vertex AI, Azure ML, CI/CD/CT, monitoring
Data Engineering: SQL, ETL/ELT pipelines, orchestration, data quality
Cloud & DevOps: AWS/Azure/GCP, Docker, Terraform
Other: Vector DBs (Pinecone, Weaviate, Chroma, Milvus), observability, model optimization, model explainability (SHAP/LIME)

----------------------------------
| Roles And Responsibilities |
----------------------------------

- Architect end-to-end ML systems covering data ingestion, feature engineering, model training, deployment, and monitoring.
- Design scalable ML pipelines for classical ML, deep learning, and GenAI workloads including LLMs, RAG, embeddings, and semantic search.
- Select optimal algorithms, frameworks, and cloud platforms to address complex business and technical challenges.
- Implement model optimization strategies such as quantization, pruning, distillation, and LoRA/QLoRA fine-tuning.
- Build LLM-powered solutions using prompt engineering, knowledge grounding, agents, and retrieval-augmented generation.
- Work with top LLMs (GPT-4/5, Llama 3/3.5, Qwen, Claude, Gemini) and integrate vector DBs like Pinecone, Weaviate, Chroma, Milvus.
- Evaluate and improve LLM and ML model performance, reduce hallucinations, and ensure security and compliance.
- Lead development of ML, NLP, NER, and deep learning models using Transformers, CNNs, XGBoost, and other advanced architectures.
- Drive data science workflows including EDA, feature engineering, statistical modeling, and explainability (SHAP/LIME).
- Architect MLOps pipelines (CI/CD/CT), enabling model versioning, monitoring, drift detection, and automated retraining.
- Deploy models at scale using Kubernetes, Docker, Terraform, serverless components, and GPU-based environments with strong observability.
- Collaborate with cross-functional teams, lead technical discussions and POCs, and mentor ML/AI teams to deliver enterprise-grade solutions.

Write "I have worked with LLMs, RAG, Embeddings" at the start of your proposal so I know you have read the description. Long term contract . (Min 1 year)

Minimum Contract Payment: $2000 per month depending on experience level.

• Describe your recent experience with similar projects
• Please list any certifications related to this project

We're a growing affiliate network looking for a skilled AI Engineer to help us build and automate intelligent systems that improve our operations, partner performance, and revenue tracking.

You'll work directly with our team to design, implement, and maintain AI-powered agents and automation workflows — from lead routing and offer matching to performance monitoring and reporting pipelines.

This is a part-time, ongoing engagement with potential to grow.

🔧 What you'll work on:
• AI agent development (LangChain, CrewAI, AutoGen, or similar)
• Automating repetitive workflows across our affiliate tech stack
• Integrating LLMs with internal tools, CRMs, and dashboards
• Building data pipelines to surface actionable insights
• Prompt engineering and optimization for business-specific tasks

✅ Requirements:
• Proven experience building AI agents or automation systems
• Strong Python skills; familiarity with LLM APIs (OpenAI, Anthropic, etc.)
• Experience with tools like n8n, Zapier, Make, or custom-built orchestrators
• Comfortable working async and independently
• US-based (required for compliance and time zone alignment)

⭐ Nice to have:
• Background in affiliate marketing, ad tech, or performance marketing
• Experience with affiliate platforms (Impact, PartnerStack, TUNE, etc.)
• Familiarity with tracking pixels, postbacks, or attribution systems

📋 To apply, please include:
• A brief description of a similar AI/automation project you've built
• Your hourly rate
• Your availability (hours per week)

We respond to every applicant. Looking forward to hearing from you!

We're classifying entities from a domain with over 200 years of continuous manufacturing history, hundreds of makers across dozens of countries, and a body of institutional knowledge that is vast, fragmented, and frequently contradictory.

Some manufacturers have been in continuous operation for over a century. Others were acquired, renamed, dissolved, revived, or split apart, sometimes more than once.

A single model name might refer to fundamentally different designs depending on the decade, the country of origin, or which generation of ownership produced it.

Lines have undergone running changes that were never formally documented. Naming conventions overlap, get recycled, and in some cases the market's name for something has no relationship to what the manufacturer called it.

A single nameplate may encompass dozens of legitimate configurations, differing in specification, era, finish, or regional market, each of which matters for identification and valuation. Critical attributes sometimes appear in a name, sometimes in a description, sometimes are only inferable from images or context, and sometimes can't be inferred at all.

Our input data comes from thousands of individual users, each describing what they have in their own words, with their own conventions, their own shorthand, and their own mistakes. There is no standardization at the point of entry.

Descriptions range from highly knowledgeable and precise to vague, incorrect, or contradictory. Some users over-specify, some under-specify, and some confidently misidentify what they have entirely. This is the raw material we work with.

On the other end, we need deep faceted classification and well-structured attribute trees, deduplicated and logically grouped, so that the same entity described a hundred different ways by a hundred different people resolves to one canonical record. We need to be able to slice across makers, nameplates, eras, specifications, and configurations in ways that are consistent, queryable, and meaningful.

We have an advanced extraction methodology already in place and a technical team building it. What we need is someone who can work alongside our data scientists to sharpen the ontology itself: defining entities precisely, writing classification rules and decision logic, testing those definitions against real-world edge cases, and refining the taxonomy when the data fights back.

We're looking for help with...

Taxonomy and Ontology Design:
Audit and refine our existing hierarchy. Establish documented rules for where canonical identity ends and descriptive attributes begin. Build authority control standards for maker names, model names, and variant designations. Design the faceted classification structure and attribute schemas so that entities are properly deduplicated and grouped across the full depth of the variant tree. Create decision frameworks for genuinely ambiguous cases.

Classification Rules and Extraction Alignment:
Work with our data science team to translate taxonomic decisions into classification schedules and decision logic that inform our extraction pipeline. Account for the reality that input data is messy, inconsistent, and often wrong. Design confidence thresholds and fallback paths. Build validation rules that catch systematic misclassification.

Ongoing Refinement:
Diagnose real-world classification failures. Determine whether the root cause is a rules problem, a schema problem, or a data quality problem. Document patterns and help us build a feedback loop between classification performance and taxonomy evolution.

Hiring Criterai:
You have depth in taxonomy design, ontology modeling, or knowledge organization, whether from an MLIS, information architecture, or years of applied cataloging and classification work. You understand faceted classification, controlled vocabularies, and authority control. You've worked in domains with real complexity: long manufacturing histories, deep variant trees, collector markets, or institutional knowledge that lives in people's heads rather than databases.

You're comfortable working alongside computational pipelines. You don't need to write code, but you understand the difference between a classification problem and an engineering problem, and you know how to write rules precise enough that a system can act on them.

Most importantly, you find this kind of problem genuinely interesting.

To Apply Please Share:
- An example of a taxonomy, controlled vocabulary, or classification schema you've designed (can be anonymized)
- How you approach cases where the data supports multiple valid classifications
- Your availability

Hi! We are a growing healthcare company with a need for some great data analytics/skills to help us continually analyze and improve upon our numbers so this means hitting KPIs each month, analyzing our opportunities and ensuring that we're hitting our goals. Forecasting experience is helpful as is general data analytics skills with Excel and coding.

I am looking for an integration and data specialist to help me set up some custom client reports and dashboards. The dashboard is being built out by my developer but I need someone to help connect the data and make sure it is accurate. Is this something you can do?

looking for someone to help me learn who is
Power Automate AI Builder expert
Microsoft 365 automation consultant
workflow automation Power Automate

I have a few projects that I would need to do as i’m learning . I want to build real workflows, not just learn basics and need help setting up automations for current business operations

Please advise if you can assist

Please don't apply if you don't have experience of these in Ubuntu based platform

Scope of Work

1. Big Data Storage & Processing
2. Database Benchmarking

Conduct a comparative performance analysis of two databases (one SQL, one NoSQL) using YCSB with multiple workloads. Capture and analyze performance metrics.

3. Streaming Use Case

Demonstrate a real-time streaming pipeline using Apache Spark Structured Streaming and/or Kafka, ingesting from a public API or user input, with relevant screenshots and result interpretation.

4. Architecture Design

Produce a clear architecture diagram of the full big data stack (Hadoop/Spark, SQL/NoSQL, programming layer) with written rationale.

5. Advanced Analytics & NLP

Perform EDA, data wrangling, and sentiment extraction from tweets (justify the chosen sentiment technique — VADER, TextBlob, FinBERT, etc.).
Merge tweet sentiment with stock price data as exogenous features.

6. Time Series Forecasting

Implement at least 2 forecasting models, including:
1 autoregressive model (ARIMA / SARIMA / SARIMAX)
1 neural network (LSTM / GRU / Transformer)
Forecast CLOSE price at 1, 3, and 5 days ahead for at least 5 companies.
Include hyperparameter tuning with justification (Grid Search, Optuna, Bayesian, etc.).

7. Interactive Dashboard

Build a dynamic, interactive single-screen dashboard (Plotly Dash, Streamlit, or Power BI) displaying all forecasts and key visuals. Design must reflect Tufte's data visualization principles.


Deliverables

Jupyter Notebooks (.ipynb) with clean, commented code (max 2 notebooks).
Processed data files and any extracted CSVs.
Technical written report (Word .docx, ~3,000 words) in standard report format with index/appendix, covering rationale, methodology, results, architecture diagram, and references in Harvard style.

We are seeking a highly skilled AI Forward Deployed Engineer for an important long-term project. The ideal candidate will have extensive experience in deploying AI solutions and will be able to discuss complex technical concepts and methodologies in depth. This interview aims to assess the candidate's expertise, problem-solving abilities, and experience in AI implementation. If you have a strong background in AI technologies and can communicate effectively, we would like to hear from you.

• Please describe the most technically complex customer deployment you've overseen? What was the architecture, what went wrong, and how did your team resolve it?
• Describe your recent experience with similar projects

Skills required:
• Node.js or Python (Python preferred — better LLM ecosystem)
• Anthropic Claude API (must have shipped at least one Claude project)
• Gmail API + OAuth
• Supabase or Postgres
• Slack/Discord bot frameworks (Bolt for Slack, [discord.py](http://discord.py/) or discord.js)
Skills NOT required (yet):
• React Native (that’s Phase 4+)
• DevOps / Kubernetes
• ML/data science

I'm building an AI-first tool for influencer talent management
(we manage creators in). Looking for a contractor to build the MVP: Slack/Discord bot + Gmail integration +
Claude API for deal extraction.

Phase 1 is ~2 weeks of work, ~15 hrs/week.
I have UI/UX already done, tech stack picked (Python, Supabase,
Claude API), and a clear scope.

Open to chatting? Happy to share the spec doc.

1. “Show me the last Claude API project you shipped
2. “How would you handle the Gmail OAuth flow for read + draft scope only?”
3. “What’s your async communication style?”
4. “What would you build first if I gave you Phase 1 tomorrow?”
• Weekly 30-min sync (Monday morning, video). Review last week’s progress, set this week’s scope.
• Daily async standup in Slack/Discord: What I shipped yesterday. What I’m shipping today. Anything blocked.
• Loom videos for handoffs, not text-only. Five minutes of screen recording beats five paragraphs.
• End-of-phase demo

We are building a geospatial market intelligence platform that aggregates, enriches, analyzes, and exposes large volumes of public and semi-public data, including government permits, regulatory filings, demographic and census data, parcel and property records, crime data, environmental data, parking and infrastructure inventory, zoning data, business activity, construction signals, and other location-based datasets.

The goal is to turn messy real-world data into a queryable, map-driven intelligence product that helps us identify market patterns, spatial relationships, development activity, business opportunities, and location-based trends.

This is not simple web scraping.
This is not email scraping, lead scraping, or collecting business names into a spreadsheet.

We need someone who can build serious data infrastructure around difficult public datasets. The right person can look at a vague target like:

"Find the public data that explains what is changing in this market, keep it fresh, and make it queryable."

...and turn that into a working pipeline.

You will be expected to identify sources, design the scraper or ingestion architecture, normalize the data, enrich it, store it properly, expose it through APIs or MCP servers, and help make it usable by AI agents, dashboards, and internal tools.

WHAT THIS ROLE COMBINES

This role requires a strong mix of:
• Senior Python and/or Node data engineering
• Advanced scraping, crawling, and ingestion architecture
• Public-data source discovery and evaluation
• Geospatial data processing
• PostGIS and spatial database design
• Entity matching, deduplication, and enrichment
• AI-agent workflow development
• MCP server development
• LLM-powered data querying and automation
• Production pipeline monitoring and reliability

We are looking for a builder who can think through unclear problems, make smart technical decisions, document those decisions, and ship working systems without needing step-by-step instructions for every task.

WHAT YOU'LL OWN

You will build and maintain ingestion systems for public and semi-public datasets across federal, state, county, and municipal sources.

Your responsibilities may include:
• Building scrapers, crawlers, ingestion bots, and API connectors for complex public data sources
• Creating custom Python, Node, Playwright, BrowserUse, Crawlee, or Apify-style automation where normal API access does not exist
• Designing reliable automation strategies for difficult public portals, including session handling, retries, queueing, rate limits, source-change detection, and monitoring
• Working with geospatial data pipelines using PostGIS, shapefiles, parcels, geocoding, spatial joins, Mapbox, Leaflet, or similar tools
• Normalizing messy data across permits, parcels, ownership records, crime datasets, demographics, infrastructure, business activity, and regulatory filings
• Performing entity matching and deduplication across inconsistent government and commercial datasets
• Creating enrichment logic that turns raw records into usable market intelligence
• Building MCP servers that expose structured data to LLMs, AI agents, internal dashboards, and automation workflows
• Building APIs that allow internal systems to query geospatial, business, parcel, and market intelligence data
• Creating AI-agent workflows that can search, reason over, and act on the data we collect
• Building monitoring, alerting, and self-healing systems so pipelines stay fresh
• Documenting what was built, why it was built that way, and how it can be maintained or extended

We do not want someone who simply downloads a CSV and calls it done.

We need someone who can help build systems that answer questions like:
• Where is development activity accelerating?
• Which areas are showing signs of commercial growth?
• How do permits, crime, property ownership, demographics, zoning, and business activity relate to each other?
• Which parcels, corridors, or trade areas are becoming more valuable?
• Which neighborhoods are underserved, oversaturated, or shifting?
• What public signals suggest a location is becoming more attractive for a specific type of business?
• What patterns are emerging across property records, infrastructure, permits, and local activity?
• How can AI agents query this data and return useful business intelligence without a human manually digging through files?

AI AGENTS, LLMS & REQUIRED TOOLING

AI tooling is not optional for this role.

You must actively use modern AI development tools in your daily engineering workflow, especially Cursor AI, Claude Code, Claude, ChatGPT/OpenAI models, or equivalent LLM tools.

We are not hiring someone who only uses AI as a fancy search bar. We are hiring someone who knows how to use AI tools to move faster, debug better, design systems, document decisions, write cleaner code, and build agentic workflows that make our internal stack smarter.

You should be able to build or work with:
• AI agents and automation bots
• Tool-calling workflows
• MCP servers
• LLM-accessible APIs
• Structured prompts and agent instructions
• Data-query agents
• Research agents
• Scraping agents
• Monitoring agents
• Cursor AI and Claude Code development workflows
• Human-in-the-loop review systems where needed

You should understand how LLMs fit into real production systems, including the difference between prompting an LLM, giving an LLM access to tools, exposing structured data through MCP, using APIs, applying RAG or vector search where appropriate, and knowing when deterministic code should do the work instead of an AI model.

Before starting, you must have active access to:
• Cursor AI
• Claude Code
• Claude
• ChatGPT, OpenAI API, or equivalent LLM access

Paid subscriptions are expected. If you do not actively use these tools in your normal development workflow, this is not the right role.

As part of your application, include a 2–3 minute Loom video showing your real workflow using Cursor AI and/or Claude Code on an actual technical task. Screenshots will be ignored. We want to see how you think, inspect code, use AI tools, reason through problems, and move from vague requirement to working implementation.

WHAT THIS IS NOT

This is not basic scraping.
This is not email scraping.
This is not collecting random business names into a spreadsheet.
This is not a one-off automation script with no production system behind it.
This is not a role where we hand you a perfect source list and step-by-step instructions.
This is not a role for someone who waits for us to explain every source, every schema, every field, and every edge case.

This role requires someone who can build around the reality of public data:
• Sites change
• Data formats change
• Public portals break
• Downloads fail
• Schemas move
• APIs disappear
• Records are inconsistent
• Government data is messy

That is part of the job. We expect you to adapt, document the change, and keep the data flowing.

YOU MUST BE ABLE TO PROVE

You must show real experience with:
• Senior-level Python and/or Node development
• Scraping and automation at scale
• Playwright, BrowserUse, Crawlee, Apify-style architectures, or similar systems
• Difficult websites, dynamic portals, pagination, downloads, changing schemas, and unreliable public data sources
• Geospatial data in production
• PostGIS, Postgres, spatial joins, shapefiles, parcels, geocoding, or spatial analysis workflows
• Strong SQL and database design
• Supabase experience is a plus
• MCP server development or a clear technical understanding of how to build MCP servers quickly
• AI-agent development or tool-calling workflows
• LLM-powered automation, code workflows, or data-query systems
• Building systems that stay running, stay fresh, and recover when sources change

You should be comfortable with messy work.

You should be able to look at a government portal, parcel dataset, GIS download site, permitting system, PDF archive, public records portal, or map-based database and figure out how to extract usable data from it.

IDEAL EXPERIENCE

Strong candidates may have experience with:
• Public data pipelines
• GIS systems
• Government open data portals
• County appraisal district data
• Permit databases
• Crime datasets
• Census and demographic data
• Parcel and property ownership records
• Zoning and land-use data
• Environmental datasets
• Parking, traffic, infrastructure, or mobility data
• Market intelligence platforms
• Real estate analytics
• Location intelligence
• Retail site-selection analytics
• Urban planning data
• Economic development data
• AI-powered internal tools
• MCP servers
• Agentic data workflows

HOW WE WORK

We move fast.
We expect proactive communication.
You update us before we ask.
You flag blockers in hours, not days.
You fill logic gaps yourself.
You make the call, document it, and move on.

We do not want someone who disappears for a week and comes back saying:
"The website changed."

We expect you to notice the website changed, adapt the scraper, update the pipeline, document the change, and keep the data flowing.

We want builders, problem-solvers, and technical thinkers who can combine data engineering, geospatial thinking, AI tooling, and production discipline.

MUST READ TO APPLY:

Secure Browser Configuration: Guided installation of Tor and Brave on Windows, ensuring advanced privacy settings, shields, and security protocols are fully optimized for data sovereignty.

Privacy Best Practices: Provide a brief walkthrough on utilizing both browsers effectively to maintain operational privacy.

Troubleshooting: Address any immediate connectivity, DNS, or performance issues that arise post-installation.

Start your application with the word:
PERMIT

It must be the first word of your message. Anything else will be archived without being read.

In your application, include:
1. The Loom video described above — non-negotiable
2. Two real projects you have shipped involving scraping, geospatial data, public datasets, MCP work, AI agents, or data pipelines — with links
3. HOW HAVE YOU USED TOR BROWSER FOR SCRAPING?
4. A list of the AI tools you actively use, including Cursor AI and Claude Code
5. How long have you had experience with Tor Browser or Brave Browser?

We reply to applications that follow these instructions. We do not reply to ones that do not. If you do not know Tor Browser or Brave Browser and scraping with them, do not apply.

• Describe your recent experience with similar projects
• Please list any certifications related to this project
• What tools do you use for data mining and visualization?

Title: Computer Vision Consultant: Football Player Tracking & Analytics (Sports Tech)
Category: AI & Machine Learning › Computer Vision
Engagement: Consulting / Advisory (hourly or fixed milestone)
Experience Level: Expert
Project Length: 1–3 months initially, with potential to extend

About Us
We’re a sports technology group building a sports tech product , a talent analytics platform that combines GPS data, video, and AI-driven assessments to predict player development trajectories. Our infrastructure runs across performance centres in Madrid and Singapore, and we’re scaling the computer vision layer of our product.

What We Need
We’re looking for an experienced Computer Vision expert to advise our internal tech team on architectural and implementation decisions. This is an advisory role, you won’t be writing production code, but you will be shaping how we build it.
Specifically, we need guidance on:
• Multi-player tracking and re-identification in match and training footage (fixed-camera and broadcast angles)
• Pose estimation and biomechanical event detection (passes, shots, first touches, defensive actions)
• Ball tracking and event tagging at scale, including handling occlusion and low-resolution footage
• Choice of model architectures (YOLO variants, ByteTrack, MMPose, custom transformers) vs. off-the-shelf APIs
• Data annotation pipeline design, we currently use a 38-tag event taxonomy and want to optimise the human-in-the-loop workflow
• Edge inference considerations (we’re piloting Meta smart glasses as a data acquisition layer)
• Build vs. buy trade-offs against providers like PFF FC, Stats Perform, and Gradient Sports
Ideal Background
• 5+ years working on production CV systems, ideally in sports, surveillance, or robotics
• Hands-on experience with player tracking, pose estimation, or action recognition
• Familiarity with sports analytics is a strong plus (experience with broadcast football data, tactical cameras, or similar)
• Comfortable reviewing code and architecture proposals from our internal engineers
• Able to translate technical decisions into clear business trade-offs (cost, accuracy, latency)
Engagement
• Initial scoping call (paid) to align on the project
• Weekly 60-minute advisory sessions with our tech lead
• Async review of architecture documents and model selection decisions
• Optional: hands-on prototyping if the fit is right

To Apply
Please share:
1. A short summary of relevant CV projects you’ve worked on (sports preferred)
2. Your hourly rate for advisory work
3. Your availability over the next 4 weeks
4. One opinionated take on the current state of player tracking in football, we want to see how you think

We are seeking a skilled Data Analysis Expert with a strong background in Python and Power BI, coupled with AI engineering experience. The ideal candidate will be responsible for analyzing complex datasets, visualizing the results, and implementing AI solutions to enhance our data processes. Strong communication skills and the ability to work collaboratively in a team environment are essential. If you are passionate about data-driven decision-making and have a knack for innovation, we want to hear from you!

We are looking for one or more programmers to work on Python-based simulation models and AI automation of those models at sistemaFutura.

At sistemaFutura, we combine simulation modeling and behavioral science to study and improve complex problems. We work on a range of issues, from gender-based violence to climate adaptation and financial inequality; in each case, the concept is the same: we build empirically calibrated, theoretically guided models of a complex problem, which we analyze for opportunities to make a meaningful change.

We have already built a custom simulation engine in Python as part of a deterministic workflow with optional (user-triggered) AI-automation. We are looking for programmers to work both on the engine and the AI-automation. The key parts of the tech stack are: Python, React + Tailwind, LangChain/LangGraph, Weaviate + Temporal.

Very few candidates will have built this type of AI-enabled simulation model before -- that's fine. We are looking for a strong foundation of (A) object-oriented python skills and (B) experience in scientific modeling or complex data analyses. This generally means a mid level to senior software engineer (4-10 years of experience).

We use Claude Code and Cursor, and we'd expect you to as well, whenever useful. However, that isn't a free pass: we'll assess your ability to critically analyze the code these tools develop, critique and design your own architectures, etc.

Please note that it is very important that (a) you're in a time zone or have a work schedule that makes it easy to coordinate with the Western Hemisphere (Central US / Central America), and that (b) you are proficient at speaking English. Since we'll need to train you on some of the aspects of what we do, ease of communication (time and language) are essential.

At this point, we can't offer a full time job, but we hope to build to that over time for one or more excellent candidates.


Finally -- please please don't spam us with messages outside of Upwork. It's really annoying (and, against the terms of use)!

• What is the most complex (length of code, complexity of task) application or application component project you've personally developed?
• What are three design patterns or software patterns you've used?
• How do you approach maintaining and refactoring a large codebase to ensure code quality and scalability?
• Include a link to your GitHub profile and/or website